. Classification and prediction of clinical Alzheimer's diagnosis based on plasma signaling proteins. Nat Med. 2007 Nov;13(11):1359-62. PubMed.

Recommends

Please login to recommend the paper.

Comments

Make a Comment

To make a comment you must login or register.

Comments on Primary Papers and News

  1. Are Blood Samples Being Archived for Optimal Subsequent Utility?
    Anne Fagan, John Trojanowski, and Eric Blalock have commented on the excellence and the caveats of the study by Ray et al. Rather than expound on what my colleagues have already covered well, I would like to pick up on one comment by Tony Wyss-Coray that is crucial for studies of blood biomarkers. He is quoted as having said “many clinics do not collect blood for plasma, but instead freeze it or use it for serum.” He said that “plasma must be immediately prepared from the serum and then frozen.”

    This comment addresses a major issue with the collection and processing of blood samples for archiving and further study. Consider the major components of blood as serum, buffy coat, and erythrocytes. Wyss-Coray has commented on the fact that many clinics do not treat serum in a way that is appropriate for plasma, and he suggests a procedure that would be compatible with his method. Similarly, we have found that both protein and message extracted from cells contained in frozen buffy coat samples archived at several centers are badly degraded to the point of being useless. We have developed methods for collecting and processing leukocytes that preserve their utility for studies of gene expression (Mhyre et al., in preparation). Others have described alternate methods for RNA extraction (e.g., Kruhoffer et al., 2007; Kim et al., 2007; Marteau et al., 2005; Tang et al., 2003, and Whitney et al., 2003). Similarly, Rogers et al. (2006) have described methods for preparation of blood samples that yield material suitable for assessment of competence of Aβ binding to erythrocytes—which was found to be modulated by disease state. Although the erythrocyte fraction is frequently overlooked, this paper suggests it is also worthy of attention.

    The bottom line is that routine ways in which blood samples are collected and processed for future study are often inapplicable to important methods. In the design of future studies, attention must be given to recent developments in collecting, processing, storing, and analyzing blood samples.

    References:

    . Effects of storage, RNA extraction, genechip type, and donor sex on gene expression profiling of human whole blood. Clin Chem. 2007 Jun;53(6):1038-45. PubMed.

    . Isolation of microarray-grade total RNA, microRNA, and DNA from a single PAXgene blood RNA tube. J Mol Diagn. 2007 Sep;9(4):452-8. PubMed.

    . Collection and storage of human blood cells for mRNA expression profiling: a 15-month stability study. Clin Chem. 2005 Jul;51(7):1250-2. PubMed.

    . Peripheral clearance of amyloid beta peptide by complement C3-dependent adherence to erythrocytes. Neurobiol Aging. 2006 Dec;27(12):1733-9. PubMed.

    . Blood gene expression profiling of neurologic diseases: a pilot microarray study. Arch Neurol. 2005 Feb;62(2):210-5. PubMed.

    . Individuality and variation in gene expression patterns in human blood. Proc Natl Acad Sci U S A. 2003 Feb 18;100(4):1896-901. PubMed.

  2. The recent Nature Medicine paper by Ray, Britschgi, Wyss-Coray, and colleagues is the culmination of a series of elegant and rigorous proteomics experiments designed to identify possible biomarkers of AD in blood. As those who follow the AD biomarkers literature know all too well, the search for plasma/blood markers has yielded few, if any, viable candidates. Whether the paucity of promising candidates reflects true disease biology or methodological challenges and limitations remains to be determined. However, the present study provides compelling evidence for changes in a variety of interesting signaling molecules in blood that correlates with the clinical disease phenotype (i.e., probable AD dementia). Although the percent agreement with clinical diagnosis does not reach 100 percent with the panel of 18 markers, one would not expect it to perfectly discriminate the groups due to the known presence of preclinical AD pathology in non-demented elderly individuals (Morris and Price, 2001) and the difficulties in providing an accurate clinical diagnosis, especially at the early stages of the disease. In fact, the authors rightly avoid the terms “sensitivity,” “specificity,” or “accuracy” (which imply autopsy-confirmed disease diagnosis) when describing the performance of the biomarker panel, and instead use the terms “percent agreement with clinical diagnosis.” Potential comorbidities, medication use, condition of plasma samples (e.g., degree of hemolysis), etc., may also contribute to less than 100 percent agreement with clinical diagnosis. It would be interesting to know whether the samples that fall outside the appropriate clinical clusters are different from other samples in such respects. The inclusion of a clinical group with rheumatoid arthritis is an excellent control given the number of immune- and hematopoiesis-related proteins in their panel.

    The plasma panel of biomarkers also appears to perform well in predicting which subjects with MCI will go on to develop probable AD dementia, an issue that has clear therapeutic implications. Since substantial AD pathology, including plaques, tangles, and neuron/synapse loss, as well as cognitive impairment, is already apparent in the MCI stage in individuals who go on to develop AD dementia (Morris et al., 2001; Markesbery et al., 2006), an even more pressing issue is to identify markers that predict which individuals will go on to develop AD dementia but do so while they are still cognitively normal, i.e., prior to substantial neuron/synapse loss. The ratio of CSF tau/Aβ42 has been shown to be promising in this regard (Fagan et al., 2007; Li et al., 2007). It will be interesting to know how the current plasma panel performs in such cohorts of non-demented individuals who are clinically followed over years, or in individuals (especially those who are cognitively normal) with known amyloid pathology (Klunk et al., 2004; Fagan et al., 2006; Mintun et al., 2006).

    Although the accuracy of the plasma panel for discriminating clinical groups may not be greater than some of the more “tried and true” CSF biomarkers (including Aβ42, tau and ptau, and their ratios) (Galasko et al., 1998; Kanai et al., 1998; Andreasen et al., 1999; Sunderland et al., 2003; Fagan et al., 2007), the fact that a set of proteins in blood could yield such good discrimination of clinical groups is remarkable and incredibly promising in terms of possible future clinical application. I eagerly await the results of the next generation of experiments by Wyss-Coray and colleagues.

    View all comments by Anne Fagan
  3. This study is impressive. It was conducted by a sophisticated group familiar with the Alzheimer disease (AD) biomarker field, who report on a panel of blood chemicals that may be able to distinguish those with AD from normal individuals. These analytes may also identify those persons at increased risk of developing AD. As such, this is a very important study. What’s needed now is confirmation of the present findings in a larger sample of patients and controls, including comparisons of people who have dementia due to AD with people whose dementia is caused by other mechanisms.

    It also is important to follow as many living subjects as possible to autopsy to confirm the clinical diagnoses, as has been done for CSF tau and Aβ, two of the AD biomarkers most extensively studied to date. This is no small task. It has taken more than a decade, and studies of thousands of living AD patients, controls, and other dementia subjects, as well as autopsy studies on more than a hundred patients and controls to confirm and validate the potential utility of CSF tau and Aβ as AD biomarkers (1,2,6,7). This is critical for the candidate biomarkers reported by Ray et al., since the analytes they identified were not among those that were selected as being the most promising AD biomarkers by an AD Biological Markers Working Group in 2003 (2). Most are also not among those that have been reported in more recent proteomic studies of plasma from AD patients (3,4). It is well known that many initially promising biomarkers of AD have not stood up to further testing along the lines suggested above (2), so additional follow-up studies are needed before one can judge the utility of the plasma analytes reported by Ray et al. for the diagnosis of AD. However, the rigor of the studies reported here offers promise that the analytes the investigators identified may have staying power.

    Further, it is important to emphasize that AD biomarkers can have different uses. These include identifying those at greatest risk to develop AD, confirming the diagnosis of AD, epidemiological screening, predictive testing, monitoring progression and response to treatment, enriching clinical trials with specific subsets of patients or at-risk individuals, studying brain-behavior relations. Not all AD biomarkers are likely to be informative for each of these clinical and research applications, and some that are suitable to aid in clinical diagnosis may not be useful for monitoring responses of AD patients to therapeutic interventions (7). As initially proposed by the NIA/Reagan Working Group on Biological Markers of Alzheimer’s Disease (6), ideal AD biomarkers should be 1) linked to fundamental features of AD neuropathology; 2) validated in neuropathologically confirmed AD cases; 3) able to detect AD early in its course and distinguish it from other dementias; 4) reliable, non-invasive, simple to perform, and inexpensive. That working group also recommended that AD biomarkers should be evaluated for their sensitivity, specificity, prior probability, positive predictive value, and negative predictive value (6).

    The findings reported here are very exciting, and meet some of the criteria mentioned above, but it may take several years to do the studies needed before we could take these analytes to the stage where CSF tau and Aβ measures currently are for use as AD biomarkers. Indeed, CSF tau and Aβ are the standards for the AD biomarker field against which any new biomarkers should be compared. That said, this report certainly will stimulate interest in confirming and extending these studies. There are many opportunities for doing this in a timely and effective manner, including partnering with the Alzheimer’s Disease Neuroimaging Initiative, which is an innovative study supported by a public/private partnership to identify, standardize, and validate neuroimaging and chemical biomarkers for the diagnosis of AD and assessing the risk of developing AD (5).

    View all comments by John Trojanowski
  4. Dr. Wyss-Coray and colleagues investigated cytokine protein expression profiles from clinical blood samples of control, MCI, and AD subjects. Of the 120 protein species that were detectable, 18 were selected based on bioinformatic classification analyses as being highly discriminant for AD versus control subjects. There are three critical findings here:

    1. the assumption that cytokine-inflammation perturbations reported by others in AD brain tissue are reflected (albeit in a seemingly distorted manner) in blood appears to be supported;

    2. the MCI group can be subdivided by the same 18-protein panel into subjects that do or do not convert to AD; and

    3. this panel also discriminated AD from other degenerative diseases.

    Of course, there are caveats to be considered. Although these results are encouraging, before this panel can be anointed a diagnostic test, it will need testing across a much larger sample of the population. Mechanistically, it is intriguing to note that many of the inflammatory signatures reported to be upregulated in neural tissue with AD in prior studies are shown to be downregulated in blood by the present work, and this deserves further investigation. Taken together, these findings suggest that this panel may become a blood-based diagnostic test that will allow clinicians to initiate treatment in early MCI converters. In conjunction with newer treatments that appear to reduce the rate of decline in AD, this panel may become a vital component of early detection and treatment in AD.

    View all comments by Eric Blalock
  5. Wyss-Coray et al. report an investigational diagnostic test for AD that could represent a major advance for the field if their results can be replicated. A method of improving the diagnostic accuracy of AD would be valuable immediately to ensure that patients enrolled in clinical trials indeed have AD pathology. Therapeutics now in development that target Aβ are likely to be effective only in those patients with amyloid pathology (1). Additionally, while a clinical diagnosis of MCI does have predictive value for later progression to AD, a more accurate method to determine the presence or absence of amyloid pathology would facilitate the development of disease-modifying drugs for this group of patients. Diagnosis of AD versus other dementias and the identification of MCI patients who progress to manifest AD are addressed in this paper.

    The 18 signaling proteins found in this study are of interest not only as a collective diagnostic marker, but also as clues to understanding the pathophysiology of the disease. A point to consider is that the 120 known signaling proteins initially screened were chosen based in part on the availability of an assay to quantify each of them simultaneously. Genetic studies can be based on evaluations of candidate genes or based on an agnostic genome-wide screen using SNPs; similarly, proteomic studies may examine candidate proteins or use a more agnostic approach using mass spectroscopy. The proteomics approach used in this study is broader than an analysis of specific candidate proteins, but is somewhat limited in that it focuses only on the signaling proteins that were available for the described filter-based arrayed sandwich ELISA. Most of the 18 signaling proteins identified in this study have not been discussed previously as potential biochemical biomarkers for AD (2-4); while this does not necessarily detract from their utility as diagnostic markers, they are likely to reflect only a portion of the pathophysiology of AD.

    This study very nicely demonstrates the accuracy of this diagnostic method when comparing AD patients to control subjects and to patients with other neurological conditions. A limitation in the vast majority of studies of diagnostic tests is that the “gold standard” is based on a clinical rather than autopsy diagnosis. The use of a training set of samples followed by a test set to some degree addresses this limitation, since identical confounding variables in two disparate cohorts would be relatively unlikely. Additionally, the authors report results for a small number of patients for whom the diagnosis was confirmed at autopsy; the plasma markers correctly classified eight of nine individuals with AD pathology and 10 of 11 subjects with other causes of dementia. For patients with MCI, even with a relatively short follow-up period of 2-6 years, the results from this study appear promising.

    The utility of this technique will also need to be compared with other imaging and CSF diagnostic techniques now in development (4-8). The relative ease of using a blood sample makes this technique attractive—if this method were used to screen large numbers of people, either with clinical symptoms or at-risk, a staged approach with imaging or CSF analysis following a positive plasma test could be considered. Relative costs of these diagnostic modalities will also be an important consideration.

    The ultimate utility of this technique will be determined by replication in other cohorts. The Alzheimer’s Disease Neuro Imaging (ADNI) study is a large longitudinal observational study of research subjects with AD, MCI, or normal aging (9,10). A number of biochemical and imaging biomarkers are incorporated in the study such that cross-sectional and longitudinal data are obtained for subjects in each of the three diagnostic cohorts. Imaging techniques employed in ADNI include volumetric MRI, FDG-PET, and PIB; biochemical biomarkers evaluated in plasma and CSF include Aβ, tau, and isoprostanes; clinical data include ADAS-cog scores and a neuropsychological battery. Data from ADNI are available to any qualified investigator via the ADNI website. Additionally, physiologic fluid specimens (plasma, CSF, and urine) can be obtained with appropriate approval. The plasma samples obtained from ADNI could provide one source of replication for the diagnostic test described by Wyss-Coray et al.

    See also:

    Fletcher PT, Wang AY, Tasdizen T, Chen K, Jagust W, Koeppe R, Reiman E, Weiner MW, Minoshima S, Foster NL. Variability of normal cerebral glucose metabolism from the Alzheimer’s Disease Neuroimaging Initiative (ADNI): Implications for clinical trials. Ann Neurol (in press).

    View all comments by Eric Siemers